Public Good Experiments in the Lab and in the Field

Florian Englmaier & Georg Gebhardt LMU München

Public Good Experiments: Set Up A Typical Public Good Experiment • Subjects play in groups of four • Subjects get 20 tokens and must decide how many tokens they put in a common pool • Tokens in the pool are multiplied by 1.6 and distributed among the group members • Tokens outside the pool belong to the subject • This is repeated ten times with the same groups (partner treatment) March 16, 2009

TSE: „Public Goods …“

2

Public Good Experiments: Analysis

• 20 into the pool is efficient • 0 into the pool is individually rational Æ Social Dilemma

March 16, 2009

TSE: „Public Goods …“

3

PUBLIC GOOD EXPERIMENTS: RESULTS

(Data from our own lab study)

March 16, 2009

TSE: „Public Goods …“

4

PUBLIC GOOD EXPERIMENTS: RESULTS Results from our own lab study, but representative of general results: – Even initially contribution not 20 – Contribution level goes down over periods – Some people contribute more than others

March 16, 2009

TSE: „Public Goods …“

5

PUBLIC GOOD EXPERIMENTS: ROBUSTNESS •

Many variations of this experiment exist: – Classic: • Kim & Walker (1984), Isaac et al (1985)

– Per capita return • Kim (1984), Isaac & Walker (1988)

– Symmetry/Homogeneity • Return: • Preferences:

Palfrey & Rosenthal (1991), Palfrey & Prisbrey (1993, 1997) Fischbacher & Gächter (2008), Gächter & Thöni (2005)

– Partner/Stranger • Andreoni (1988)

– Communication • Isaac & Walker (1991)

– Anonymity • Fehr and Gächter (199?)

– Punishment • Fehr and Gächter (2000)

– Learning • Andreoni (1988)

March 16, 2009

TSE: „Public Goods …“

6

PUBLIC GOOD EXPERIMENTS: VALIDITY

Why do we care so much?

March 16, 2009

TSE: „Public Goods …“

7

PUBLIC GOOD EXPERIMENTS: VALIDITY • The game serves as a metaphor for many economics situations: – classic public goods (defence, public infrastructure, health system, environmental protection, …) – intellectual property rights for informational goods – team production in organizations –… March 16, 2009

TSE: „Public Goods …“

8

PUBLIC GOOD EXPERIMENTS: VALIDITY • The game serves as a metaphor for many economics situations: – – – –

classic public goods intellectual property rights for informational goods team production in organizations …

• So it would be nice to see whether we can learn something from our lab results for behavior „in the wild“. March 16, 2009

TSE: „Public Goods …“

9

APPLICATION: Environmental Economics PG Problems are very prominent in Environmental Economics: - Slowing climate change - Reducing carbon emissions - Reducing air pollution - … Similar Structure: Common Pool Problems - Over-fishing, … March 16, 2009

TSE: „Public Goods …“

10

APPLICATION: Environmental Economics Question: How can we overcome the market failure caused by externalities? If we know that behavior in the field and in the lab are related (and that the results are transferable to a whole class of structurally related problems) we can start to analyze solutions to the problems in the lab. March 16, 2009

TSE: „Public Goods …“

11

Our Study – Time Line • First we run a field experiment – Real effort task; participants are not aware that this is an experiment; three treatments – one resembles PG

• We invite participants to the lab – Play a „typical“ PG experiment

• We compare behavior in the lab and in the field (within subject) March 16, 2009

TSE: „Public Goods …“

12

Field Experiment: Task • 103 Subjects recruited to register of books in our library • Applicants had to complete a test on the recruiting website Æ ability measure • Salary € 55, job duration „up to 5 hours“ • Groups of four, two shifts (morning, afternoon) • Welcome and introduction together (by student helper) • Work separately in four adjacent offices on laptops with Excel • Student helper brings books and gives feedback every 30 min • At the end, one of us pays them money and invites them to a laboratory experiment March 16, 2009

TSE: „Public Goods …“

13

Field Experiment: Treatments Three treatments • Single: Each worker can leave once 170 books have been registered - Feedback on individual progress • Group: All workers can leave once 680 books have been registered - Feedback on group progress • NoIncentives: All workers must work 3.5 hours - Feedback on individual progress March 16, 2009

TSE: „Public Goods …“

14

Field Experiment: Set Up

March 16, 2009

TSE: „Public Goods …“

15

Field Experiment: Set Up

March 16, 2009

TSE: „Public Goods …“

16

Field Experiment: Set Up

March 16, 2009

TSE: „Public Goods …“

17

Public Good vs. Field Experiment Game theoretic structure similar • Full marginal return in Single treatment. • Only a share of marginal return in Group treatment. Prediction: March 16, 2009

Inefficiently low contribution TSE: „Public Goods …“

18

Public Good vs. Field Experiment But many differences: • • • •

Effort in, time out vs. tokens in, tokens out Dynamic vs. static Labor market vs. neutral framing Real life vs. lab

March 16, 2009

TSE: „Public Goods …“

19

Field Experiment: Results • Descriptive Statistics – Group: • Mean 172,41; StdDev. 32,40; # Participants 44

– Single 169,00 7,29 • Mean 169,00; StdDev. 7,29; # Participants 32

– NoIncentives • Mean 183,04; StdDev. 44,28; # Participants 27

– # males 30; # females 73 – Avg Work Time » Group 191.61 min » Single 195.2 min » NoInc 207.5 min March 16, 2009

TSE: „Public Goods …“

20

Field Experiment: Results

Baseline: NoIncentives Standard errors in parenthesis, clustered at the session level * 5%; ** 1%; *** 0.1% significance March 16, 2009

TSE: „Public Goods …“

21

Field Experiment: Results Performance NoIncentives < Performance Group < Performance Single

March 16, 2009

TSE: „Public Goods …“

22

Field Experiment: Results Performance NoIncentives < Performance Group < Performance Single

Within Group: Standard deviation of the online test has an effect: Æ Members of more heterogeneous teams do worse!

March 16, 2009

TSE: „Public Goods …“

23

Public Good Experiment

• About 2 weeks after the field experiment is over, we run 5 sessions at MELESSA in Munich • Roughly half of our subjects from the field come to the lab (49 subjects) • We fill the sessions with people from the MELESSA subject pool March 16, 2009

TSE: „Public Goods …“

24

Public Good Experiment

• In total we have 84 subjects (4*16 and 1*20) in the lab. • Each session lasts roughly 90 minutes. • Avg. Earnings: 23,85€

March 16, 2009

TSE: „Public Goods …“

25

Public Good Experiment: Design •

Experiment has 6 rounds – Rounds 1-3 non fully anonymous – Rounds 4-6 fully anonymous – 1. PG , 2. PD, 3. PD, 4. PG, 5. PD, 6. PD

• •

PG = Standard Public Good as described previously PD = Public Debt; new design (not covered in talk; results very similar)

March 16, 2009

TSE: „Public Goods …“

26

Public Good: Decision Screen

March 16, 2009

TSE: „Public Goods …“

27

Public Good: Results by Period

March 16, 2009

TSE: „Public Goods …“

28

Public Good: Results

Standard errors in parenthesis; * 5%; ** 1%; *** 0.1% significance March 16, 2009

TSE: „Public Goods …“

29

Public Good: Results Merged

March 16, 2009

TSE: „Public Goods …“

30

Exit Survey We ask in total 14 questions from the World Value Survey and the SOEP – – – – – –

Q1_trust Q2_freedom Q3_fairness Q4_PC Q5_coscientiousness Q6_risk • • • • • •

Q7a_risk_car Q7b_risk_money Q7c_risk_sport Q7d_risk_carreer Q7e_risk_health Q7f_risk_strangers

– Q8_satisfaction – Q9_satisfaction_5yr March 16, 2009

TSE: „Public Goods …“

31

Public Good: Relation to Survey No clear pattern; nothing really seems to matter.

Standard errors beneath coefficient; * 5%; ** 1%; *** 0.1% significance March 16, 2009

TSE: „Public Goods …“

32

Relation Public Good & Field Experiment

Now, is there a relation between the lab and the field?

March 16, 2009

TSE: „Public Goods …“

33

Relation Public Good & Field Experiment

March 16, 2009

TSE: „Public Goods …“

34

Relation Public Good & Field Experiment Dependent Variable: Average Contribution Field

Standard errors in parenthesis, clustered at the session level * 5%; ** 1%; *** 0.1% significance March 16, 2009

TSE: „Public Goods …“

35

Relation Public Good & Field Experiment Dependent Variable: Average Contribution Field

The effect is large! •If avg PG contribution moves from 0-20 (min – max) this more than doubles the constant (1.38) in the field regression. •If avg PG contribution moves from the mean (9) to the max this adds +50%. •If avg PG contribution moves one SD (4.3) up this adds 25% of the constant. March 16, 2009

TSE: „Public Goods …“

36

Relation Public Good & Field Experiment Cross-Equation Tests

March 16, 2009

TSE: „Public Goods …“

37

Relation P.G. & Field Experiment

March 16, 2009

TSE: „Public Goods …“

38

Conclusion • „Fair“ behavior in Lab and Field correlated • In particular true for Public Good treatment • Indication for external validity of public good experiments • Game theoretic structure captures important aspects of real life March 16, 2009

TSE: „Public Goods …“

39

Conclusion

Thank you very much for your attention and your comments!

March 16, 2009

TSE: „Public Goods …“

40